Haswell Delidded

SonDa5

SonDa5

Supreme [H]ardness
Joined
Aug 20, 2008
Messages
7,437
I got out of The LID.


THEELID.jpg




Into the light.


4770kseesthelight.jpg



Pure Razor IHS removal.



SonDa5 said:
Here are my results for my maximum performance over clock on my 4770k before I delid it.

Powered by Intel's latest AVX2 linpack code.

Look at the Gflop performance and the heat.

4770k with HT off at 4.4GHZ.
Custom water loop, not delidded yet
Ambient temps 24C

AVX2HToff.jpg


http://valid.canardpc.com/2859272


This chip is worthy and ready for delidding.
Click to expand...



Delidded no IHS direct core mount.
Same water cooling as before.
Same BIOS settings as before.
Coolaboratory Liquid Pro TIM between DIE and Block.


4770KDelidGO.jpg


http://valid.canardpc.com/2865214

Process Gflop performance is up!
With ambient temps taken into consideration temps dropped at least 29C per core!
Also important to note that during the before delid test a 100C TDP heat was topping out and the CPU slows down a little by itself to keep from heating up. So max operation temps were never seen for the before delid test.


Conclusion:

Delidding was a big win for my hot 4770k.


Currently gaming at 4.7GHZ and temps max out at 50C range on each core.
 
Last edited:
For those that have delidded please post your before delid and after delid experiences
 
30°C + temp improvement? check.
Delidded? (not re-lidded) check.
Direct die to waterblock with liquid metal TIM? check.
...[H] so far...
Reported resulting increase in stable overclock? Still waiting.
 
okashira said:
Reported resulting increase in stable overclock? Still waiting.
Click to expand...

Working on it but greatest reward of the delid is drop in temps.

Still learning how to over clock my new Gigabyte Z87X-OC mother board so I think my max over clocks will improve once I learn the BIOS.


Before I could not game at 4.7GHZ with any reasonable gaming temps or stability. Now I'm gaming at 4.7GHZ and temps are in 50C range when gaming. Voltage is at 1.38v for stable 4.7GHZ.

Max temps under the Linpack with AVX2 code observed in low 90C on 1 core at 1.38v for 4.7GHZ. Acceptable temps and still have some room to push that benchmark. 3d benchmarks temps are way lower than this. I'm hoping to be able to pull off a 3dmark11 run with 4770k at 5GHZ.
 
Grats on the result, you seem to have a pretty good sample being able to run 4.4 at only around 1.2V

Keep in mind you would have reseated your block which you technically never eliminated before the De-lid , or at least was my impression. Temps look about 3-4C lower than my undelided [email protected] now at 1.33V
 
Last edited:
Rattle said:
Why is HT off
Click to expand...


Turned HT off as recommended by Intel code writers for the newest Linpack with AVX2.


Block and radiator is in my sig. Not sure what the water temp was but the loop cools a GPU as well. Ambient temps were listed.


Light weight copper block DT Sniper block.

DTSNIPERSIDEDirECTDiEmOUNT3570k.jpg


using a EK TRUE 115X back plate for stability.

EKS155TRUEbackplateDTSNIPER.jpg
 
Last edited:
You just convinced me that I need to de-lid. Impressive results !

Are there CLC's that work/support Socket 115x and delidded CPU's lower contact surface ?
I'm thinking about the Cooler Master Seidon 240M or the Silverstone Tundra TD02, depending on if that last one fits the Ncase M1.
 
Phuncz said:
You just convinced me that I need to de-lid. Impressive results !

Are there CLC's that work/support Socket 115x and delidded CPU's lower contact surface ?
I'm thinking about the Cooler Master Seidon 240M or the Silverstone Tundra TD02, depending on if that last one fits the Ncase M1.
Click to expand...

Thanks for the compliments. :)

I wouldn't put a block with a motor built into it directly on a bare die. That vibration directly on the CPU can't be good for it.


That small case is nice but it doesn't look like it can fit a decent loop in it.

If you are going to water cool focus on your cooling gear then figure out what case you want to house it all. Also you could possibly use the Ncase M1 if you have your radiator and pump on the outside of the case.



Good luck.
 
Rattle said:
Why is HT off
Click to expand...

HT is off because the new intel pack is able to stress the real physical cores 100%. Having HT on doesn't do jack if all the resources being used by the physical cores are not available. That and the program bugs out and uses 2 cores and 2 hyperthreads instead of stressing out the entire cpu.

You can opt to set affinity to cores 0 2 4 and 6 instead of turning HT off. It does the same thing.
 
Was mounting bare die finicky? Did you have to compensate for the space difference from not having the IHS in place when mounting your heatsink/waterblock? How did you torque your waterblock on and not worry about crushing the die?
 
Dizon248 said:
Was mounting bare die finicky? Did you have to compensate for the space difference from not having the IHS in place when mounting your heatsink/waterblock? How did you torque your waterblock on and not worry about crushing the die?
Click to expand...




I've done this at least 10 times with the same mount on Z77 and now Z87. Nice even hand snug pressure. Not tight. Its a gentle mount. Works well when done right.

My block fits fine with the IHS off.
 
And how it would work with air coolers? They are bit heavier, so I'm not so convinced that you could use bare die on the biggest things like True Copper and such? I remember old time AMD chips, that you could crush the core by just bad mount of cooler.
 
MorgothPl said:
And how it would work with air coolers? They are bit heavier, so I'm not so convinced that you could use bare die on the biggest things like True Copper and such? I remember old time AMD chips, that you could crush the core by just bad mount of cooler.
Click to expand...



I would only use bare die with light weight water block.
 
SonDa5 said:
Thanks for the compliments. :)

I wouldn't put a block with a motor built into it directly on a bare die. That vibration directly on the CPU can't be good for it.


That small case is nice but it doesn't look like it can fit a decent loop in it.

If you are going to water cool focus on your cooling gear then figure out what case you want to house it all. Also you could possibly use the Ncase M1 if you have your radiator and pump on the outside of the case.



Good luck.
Click to expand...

MorgothPl said:
And how it would work with air coolers? They are bit heavier, so I'm not so convinced that you could use bare die on the biggest things like True Copper and such? I remember old time AMD chips, that you could crush the core by just bad mount of cooler.
Click to expand...

I'd think that goes back to the vibration concern listed above, having a direct mounted fan on the CPU. No matter how well balanced the fan motor is, there's always vibration.
 
Wonder if sticking a vapor chamber heat spreader between the CPU and the water block would help even more... Tried to find a VCHS a few weeks ago but couldn't find anywhere to buy just one of the appropriate size for a CPU cooler.
 
Amazing results.

Now there is no doubt that foregoing fluxless solder and using paste was a short-sighted greedy dipshit move. Way to go, Intel. :rolleyes:
 
I find this Process very scary its always very impressive when someone delids one of these good work man
 
SonDa5 said:
I wouldn't put a block with a motor built into it directly on a bare die. That vibration directly on the CPU can't be good for it.
Click to expand...
Hmm that changes things. I'm not in the mood for a full-blown self-built WC-setup this time and the compactness of the CLC-solutions also appeals to me as space for things like a pump is at a premium in the Ncase. While I don't doubt it's cooling capacity, I doubt I'll be happy fudging more tubing and a seperate pump somewhere in that 14L case.

I wasn't aware that the vibration of the pump was on issue with bare cores.
 
using paste was a short-sighted greedy dipshit move. Way to go, Intel.
Click to expand...

I think it was a purposeful move intended to limit the performance of the mainstream platform so that the IB-E platform could have a chance.
 
drescherjm said:
I think it was a purposeful move intended to limit the performance of the mainstream platform so that the IB-E platform could have a chance.
Click to expand...

No, it was greed and stupidity. If they wanted to inhibit it in order to make the LGA2011 platform look better, they could have just implemented a multipler limit for the 1155/1150 K parts.
 
Good stuff sonda but this is nothing new direct die with IHS delidded has been around for a long time now and the results speak for themselves. It's very necessary sometimes. But Haswell is such a beast it's almost required actually.
 
Well I don't have a Haswell but I did just delid my 3770 yesterday, used CL ultra on the die and top of ihs, and the temps dropped by over 10 degrees while gaming, and it seems the harder I push it the better it performs. If you are on the fence about delidding, I recommend it. Just take your time and don't rush it. Be patient and it will pay off.
 
any gotchas to direct die cooling the core on these? Are the cores super soft or are they fairly hard like the old athalon XP days?
 
it's bare silicone. It will chip easily. stay away if you're clumsy.

The old bare die chips were often epoxy coated which really helps to protect them. But they didn't have near the W/m^2 output so I prefer the current arrangement (despite people who think intel's solution was "greedy," I call it a sound business solution that still benefits enthusiasts.)
 
Can you try to put the lid/IHS back with a good TIM and post the results ? I'm thinking of something like:
1. Remove the IHS & clean and lap it.
2. Put the IHS back on the CPU with some good TIM like liquid metal or MX4.
3. Use epoxy or sealant to glue the lid back & clamp it for 24 hrs.
That way, we can use heavy HSF on the CPU with better temp without crushing the die.
 
okashira said:
(despite people who think intel's solution was "greedy," I call it a sound business solution that still benefits enthusiasts.)
Click to expand...

:rolleyes:

If it was such a sound and beneficial decision that Intel made, then de-lidding would be as common and easy for enthusuasts as buying and applying some sort of aftermarket cooler.
 
David-Duc said:
Can you try to put the lid/IHS back with a good TIM and post the results ? I'm thinking of something like:
1. Remove the IHS & clean and lap it.
2. Put the IHS back on the CPU with some good TIM like liquid metal or MX4.
3. Use epoxy or sealant to glue the lid back & clamp it for 24 hrs.
That way, we can use heavy HSF on the CPU with better temp without crushing the die.
Click to expand...

From the results I've seen you don't need to glue the IHS back on. You remove the IHS, clean it, put some good TIM on (very sparingly) then place the cpu in the socket and use the clamp to hold the IHS onto the cpu. From the test's I've seen people have been very successful reducing temps. Check out some how-to's for specifics.
 
Back in the day we used shims to protect the die-

shims.gif


Is there such a product for current-gen processors?
 
DejaWiz said:
:rolleyes:

If it was such a sound and beneficial decision that Intel made, then de-lidding would be as common and easy for enthusuasts as buying and applying some sort of aftermarket cooler.
Click to expand...
why? I don't see the connection. Spell it out for me.
 
marshac said:
Back in the day we used shims to protect the die-

shims.gif


Is there such a product for current-gen processors?
Click to expand...

I haven't seen anyone produce these, but several forum members who do CNC work for a living have stated they were interested in it..Cmaki4 here would be a good guy to PM, he has his own metal working business, and makes custom cases etc..He might be able to work out something for you...

Honestly, I de-lidded my 3770K over a year ago, and have used both a heavy full cover Enzotech WB, and a lightweight XSPC raystorm block with great results..It simply requires a tiny bit of common sense and patience..

Here is a quick/easy way of doing this:

1. Place the cpu in the socket, spread your CL Ultra/Pro on the die (I use a cheap watercolor brush for this to ensure a nice even coat),

2. Mount your block..hand tighten the screws down lightly in a proper X pattern to ensure even torque..Start light..

3. Attempt to boot the PC..if it doesn't boot, you don't have enough pressure for the cpu to make proper contact with the pins..Tighten each a tad bit more, attempt to boot..

4. Rinse/Repeat until you have a proper boot, and sit back and enjoy your success!:D

SonDa5, as usual, great write up, it really helps the newbies ease their fears:p!
 
I know this is duh, but if taking the lid off yields so much better thermal performance, why does Intel do the lid in the first place?
 
pbassjunk said:
I know this is duh, but if taking the lid off yields so much better thermal performance, why does Intel do the lid in the first place?
Click to expand...

The lid helps protect the die. There is less chance of the die cracking during computer assembly or some thermal paste making contact with leads/diodes/resisters the paste shouldn't make contact with.
 
Indeed. It's basically because an overheating CPU isn't as big as a problem (thermal throttling) as a cracked core. Most PC's are OEM, fast install times are crucial for them.

In my opinion, they should sell the K-CPU's without a lid and with a shim attached. But they won't.
 
Nice work using the razor method! I killed a 3770k by being impatient and trying it that way, then I found the 'vice' method and round 2 worked nicely. It's absurd how much of a temp drop is experienced after delidding and using CLP, even my 3930k has better temps vs my old 3770k did during folding or gaming.
 
You must log in or register to reply here.
Back
Top